Creating missing value mechanisms

[Claire Lawson]

Hi,

I would have a full dataset with an outcome variable y and a two predictors x1 and x2 and 100 observations.

I want to make x1 missing completely at random (MCAR) with a probability of 0.5

Any help with the code would be greatly appreciated.

Thank you,
Claire

[Carlo Lazzaro]

Claire:
you may want to consider:

. set obs 100
Number of observations (_N) was 0, now 100.

. g y=runiform()

. g x_1=runiform()

. g x_2=runiform()

. g counter=runiform()

. replace x_1=. if counter>.5


. search mcartest

. mcartest x_1 x_2

Little's MCAR test

Number of obs       = 100
Chi-square distance = 0.6597    
Degrees of freedom  = 1
Prob > chi-square   = 0.4167

.

The last test is the outcome of the community-contributed module -mcartest- (see https://journals.sagepub.com/doi/pdf...7X1301300407); as it does not reject the null, there's no evidence that missing data mechanism is not MCAR.

[Claire Lawson]

Thnak you Carlo, very helpful